A Machine Learning Scheme for Boresight Error Prediction of the Antenna-Radome System

This paper presents a machine learning (ML) scheme to predict the boresight errors (BSE) in a range of antenna scan angles. With electric and magnetic currents calculated by the method of moments (MoM) at only a few scan angles, the training set is constructed by dividing the entire meshes into small pieces. A PointNet neural network is employed as the ML model in this paper. Numerical results show that the trained ML model has higher calculation accuracy than the physical optics (PO) method, and compared with MoM, this method saves calculation time because fewer scan angles need to be calculated.